Fluxes in central carbon metabolism of a genetically engineered, riboflavin-producing Bacillus subtilis strain were investigated in glucose-limited chemostat cultures at low (0.11 h−1) and high (0.44 h−1) dilution rates. Using a mixture of 10% [U-13C] and 90% glucose labeled at natural abundance, 13C-labeling experiments were carried out to provide additional information for metabolic flux balancing. The resulting labeling pattern in the proteinogenic amino acids were analyzed by two-dimensional [13C, 1H] nuclear magnetic resonance (NMR) spectroscopy. To account rigorously for all available data from these experiments, we developed a comprehensive isotopomer model of B. subtilis central metabolism. Using this model, intracellular carbon net and exchange fluxes were estimated on the basis of validated physiological data and biomass composition in combination with 2D NMR data from 45 individual carbon atom spectra in the amino acids. Glucose catabolism proceeded primarily via glycolysis but pentose phosphate pathway fluxes increased with increasing growth rate. Moreover, significant back fluxes from the TCA cycle to the lower part of glycolysis via the gluconeogenic PEP carboxykinase were detected. The malic enzyme reaction, in contrast, was found to be inactive. A thorough statistical analysis was performed to prove the reliability of the isotopomer balance model and the obtained results. Specifically, a χ2 test was applied to validate the model and the chi-square criterion was used to explore the sensitivity of model predictions to the experimental data. © 2001 John Wiley & Sons, Inc. Biotechnol Bioeng 76: 144–156, 2001.